Directory UMM :Data Elmu:jurnal:B:Biological Psichatry:Vol48.Issue6.2000:
NATURAL HISTORY
Historical Perspectives and Natural History of Bipolar
Disorder
Jules Angst and Robert Sellaro
A review of two centuries’ literature on the natural history
of bipolar disorder, including modern naturalistic studies
and new data from a lifelong follow-up study of 220
bipolar patients, reaches the following conclusions: the
findings of modern follow-up studies are closely compatible with those of studies conducted before the introduction of modern antidepressant and mood-stabilizing treatments. Bipolar disorder has always been highly recurrent
and considered to have a poor prognosis.
Bipolar patients who have been hospitalized spend about
20% of their lifetime from the onset of their disorder in
episodes. Fifty percent of bipolar episodes last between 2
and 7 months (median 3 months). The intervals between
the first few episodes tend to shorten; later the episodes
return at an irregular rhythm of about 0.4 episodes per
year with high interindividual variability. Switches from
mania into mild depression and from depression into
hypomania were frequently reported in the 19th century
and the first half of the 20th.
Antidepressant and antimanic drugs have to be given as
long as the natural episode lasts. Given the poor outcome
of bipolar disorders found in naturalistic follow-up studies
and our lifelong investigation, intensive antidepressant,
antimanic, and mood-stabilizing treatments are required
in most cases. Despite modern treatments the outcome into
old age is still poor, full recovery without further episodes
rare, recurrence of episodes with incomplete remission the
rule, and the development of chronicity and suicide still
frequent. Biol Psychiatry 2000;48:445– 457 © 2000 Society of Biological Psychiatry
Key Words: Bipolar disorder, natural history, course,
recurrence, outcome
Introduction
T
his article briefly reviews the natural history of bipolar
disorder, giving special weight to historical studies
before the era of antidepressants; integrates the results of
modern naturalistic follow-up studies; and from our own
findings 1) reanalyzes data from an early multicenter study
(Angst et al 1968b, 1973) and 2) includes some new data
From the Zurich University Psychiatric Hospital, Zurich, Switzerland.
Address reprint requests to Jules Angst, M.D., Zurich University Psychiatric
Hospital, Box 68, Lenggstreet 31, Zurich 8029, Switzerland.
Received January 13, 2000; revised April 13, 2000; accepted April 20, 2000.
© 2000 Society of Biological Psychiatry
from our lifelong Zurich follow-up study (Angst and
Preisig 1995a, 1995b). The article focuses mainly on
episodes and recurrence and, to a lesser extent, outcome; it
does not deal with rapid cycling and seasonal depression.
Recent reviews of the course of bipolar disorder have been
published by Lavori et al (1984), Keller (1987), Goodwin
and Jamison (1990), Coryell and Winokur (1992), Verdoux and Bourgeois (1995), Kessing et al (1998), Goldberg and Harrow (1999), Marneros (1999), and Bourgeois
and Marneros (in press).
The Concept of Bipolar Disorder
We owe the categorization of bipolar disorder as an illness
to Falret, who in 1851 and 1854 on the basis of longitudinal observations developed the entity of “folie circulaire” (circular madness), defined by manic and melancholic episodes separated by symptom-free intervals. In
1854 Baillarger used the term folie à double forme to
describe cyclic (manic–melancholic) episodes (Pichot
1995; Ritti 1879). Kraepelin called such cyclic episodes
“double attacks.” In both French diagnoses the prognosis
was considered to be “desperate, terrible and incurable”
(Bourgeois and Marneros, in press). Circular illness was
described by most authors as a recurrent condition; it
became the prototype of the larger group of periodic
psychoses embracing periodic mania, periodic melancholia, and periodic cyclic disorders (Ballet 1903; Mendel
1881; Pilcz 1901; Ziehen 1902, 1907).
The Concept of Mixed States
The history of the concept of mixed states has been
extensively studied by Marneros (in press): what we today
call “mixed states” were probably already known at the
beginning of the 19th century and named “mixtures”
(Mischungen) by Heinroth in 1818 and “middle forms”
(Mittelformen) by Griesinger (1845). Guislain (1852) gave
clear descriptions of different syndromes of mixed states.
The history of bipolar disorder by Haustgen (1995) traces
the term mixed states to J.P. Falret’s son Jules Falret
(1861).
Very influential in this field was Weygandt (1899), who
worked with Kraepelin and whose monograph distinguished three forms of mixed states: manic stupor, agitated
0006-3223/00/$20.00
PII S0006-3223(00)00909-4
446
BIOL PSYCHIATRY
2000;48:445– 457
melancholia (depression with flight of ideas and agitation),
and unproductive mania (elated mood, increased motor
activity, and inhibition of thinking). Kraepelin’s (1899)
textbook descriptions of mixed states were founded on
Weygandt’s monograph. Further progress was made by
Rehm’s monograph (1919, 113), which classified mixed
states systematically on the basis of the permutations of
the three elements that had been defined by Kraepelin:
thought disorder, mood, and psychomotor activity (identified as a, b, and c for mania and as A, B, and C for
depression).
Kraepelin’s Manic–Depressive Insanity
At the turn of the 19th century Kraepelin’s unifying
approach to the classification of mood disorders (1899)
resulted in bipolar disorders being subsumed within manic– depressive insanity (MDI), a broad group that included
single-episode and recurrent depression. Kraepelin (1913,
1183) was later himself to raise the possibility of the
heterogeneity of MDI. Unlike the French concepts, Kraepelin’s MDI had a good prognosis and did not develop
into severe dementia, although Kraepelin conceded the
existence of mild residual states after recovery from the
episodes themselves (Schwächezustände; Kraepelin 1913,
1349) and of mild fluctuations between episodes. Kraepelin considered periodicity to be unimportant for the diagnosis (Pilcz 1901). As a consequence of Kraepelin’s
unification of affective disorders, research on their course
frequently failed to distinguish between depression, mania, and bipolar disorder (Bratfos and Haug 1968; Fuller
1935; Paskind 1930; Pollock 1931a, 1931b, 1931c; Poort
1945; Rennie 1942; Tomasson 1947).
Notable contemporary authors nevertheless disagreed
with Kraepelin’s unitarian approach, and their studies of
the natural history of affective disorders maintained the
distinction between mania, depression, and bipolar disorder (Ballet 1903; Pilcz 1901; Ziehen 1902, 1907). This
data on the course of bipolar disorder collected in the 19th
century and the first half of the 20th, before the introduction of modern antidepressants and mood stabilizers, is of
special value in that it represents the disorder’s untreated
natural history.
Onset of Bipolar Disorder
The dating of the age of onset is to a certain extent
unreliable because it is usually retrospective and dependent on insecure recall. Bipolar disorder begins about 10
years earlier than recurrent depression, as shown by a
review of the literature (Angst 1988). Earlier studies
indicated a mean age of 28 to 33 years; epidemiologic and
newer clinical studies show that bipolar symptoms start
J. Angst and R. Sellaro
frequently in adolescence (Weissman et al 1988) and that
manic episodes manifest usually in the early 20s (Fogarty
et al 1994).
Periodic Mania and Switches of Polarity
There is considerable interest today in data on the course
of single and multiple episodes, which can answer questions about the psychopathology, duration, and frequency
of episodes; the syndromal stability over lifetime; and the
frequency with which initial major depression develops
into bipolar disorder. Today the switch of an episode from
depression to hypomania is often assumed to be drug
induced, but the phenomenon was already very common
as “reactive hyperthymia” before the introduction of
antidepressants.
A century ago the concept of periodic mania was well
known and the diagnosis much more frequent, quite
simply because it was applied to cases that today would be
considered bipolar disorders. For instance, an initial depressive syndrome cycling into a manic episode, although
frequently observed, was not considered an indication of
bipolarity (Mendel 1881; Ziehen 1902). Similarly, “postmelancholic reactive hyperthymia” with clear hypomanic
symptoms was compatible with the diagnosis of pure
periodic melancholia (Ziehen 1907, 26).
Mania switching into depression was likewise very
commonly reported as “reactive depression” (Ziehen
1902, 546, 554; Wernicke 1906, 355). Postmanic depression lasted a few days or a few weeks according to
Wernicke (1906, 355). Mild depression was observed
preceding or terminating manic attacks in most of the 128
manic patients studied by MacDonald (1918).
Modern naturalistic and treatment studies have also
found that mania frequently cycles into depression: the
rates of cycling observed in follow-up studies over 8
weeks vary from 17% (Tohen et al 1990) to 30% (Keller
et al 1986a). Our earlier retrospective record study of 300
manic patients (admitted between 1920 and 1970) found
that 21% of manic episodes cycled into depression, a rate
that did not change significantly during the intervening
decades (Angst 1987).
In a retrospective record study, depression switching
into hypomania was found in 29% of bipolar patients
hospitalised between 1920 and 1959 (Angst 1987).
Diagnostic Change from Depression to
Bipolar Illness
The syndromal course over lifetime has been little investigated. It was frequently assumed that mania was predominant in earlier years and depression in the second half of
life. Kinkelin (1954) followed-up 146 hospital first admis-
Bipolar Disorder: Natural History
BIOL PSYCHIATRY
2000;48:445– 457
447
Figure 1. Proportion of syndromes across
20 episodes in male subjects. BP, bipolar;
Man, mania; Dep, depression.
sions suffering from MDI (1929 –1947) until 1948 in a
study covering an average of 21.8 years of the total course
of the disorder. Of the 146 cases, 125 began with depression and 21 with mania. During the follow-up period 36
(28.8%) of the 125 depressive patients developed manic
episodes, a figure that would correspond to a diagnostic
change from depression to bipolar illness of 1.3% per year
of observation. Marneros et al (1991b) reported that the
initial diagnosis of depression remained stable in 79% of
cases over 27 years. In our preponderantly prospective
study we found a rate of diagnostic change from depression to hypomania/mania of about 1% per year (Angst and
Preisig 1995a). Coryell and colleagues’ (1995a) intensive
prospective follow-up study of 381 depressive subjects
over 10 years found that 10.2% developed into mania
(5.2%) and hypomania (5.0%), which also corresponds to
a 1% change per year of observation.
In their monograph, Marneros et al (1991a) found that
mania frequently developed into schizomania or mixed
states.
Syndromal Stability of the Course
New data from our Zurich follow-up study confirmed a
major gender difference in the psychopathology of bipolar
patients over the first 20 episodes (Angst 1978): female
subjects manifested significantly more depressed episodes
and male subjects more cyclic episodes (mania and depression), whereas pure manic and mixed episodes were
equally frequent in both genders (Figures 1 and 2). The
syndromal proportions were found to remain remarkably
stable over 20 episodes (Angst and Weis 1967), which also
means, for instance, that aging brings no increase in the
depressive component of bipolar illness.
Figure 2. Proportion of syndromes across
20 episodes in female subjects. BP, bipolar;
Man, mania; Dep, depression.
448
J. Angst and R. Sellaro
BIOL PSYCHIATRY
2000;48:445– 457
Table 1. Length of Episodes (Months) of Patient Samples
Mean
Mendel (1881)
Kraepelin (1913)
Panse (1924)
Wertham (1929)
Rennie (1942)
Kinkelin (1954)
Angst and Preisig (1995a)
Median
Q1
Q3
5– 6
3– 4
6 –7
4–6
2– 4
8 –10
3d
2
5
6–8
7
3.5a–5.8b
3.5– 8.4
4.3c
a
First episode.
Third episode.
Mean after logarithmic transformation.
d
Median of Table 2.
b
c
Natural Length of Episodes
Valuable data on the natural length of episodes were
published before the introduction of effective treatments
(Table 1).
Mendel (1881, 155) reported data on the length of
manic episodes (N 5 43), from which a median duration
of 5– 6 months (Q 1 5 3– 4 months, Q 3 5 6 –7 months)
can be computed; Mendel found only one episode which
lasted 10 –12 months and one over 1 year. Kraepelin
(1913) stressed the great variability in episode length but
estimated that most episodes lasted between 6 and 8
months.
Panse’s follow-up study (1924) of 205 hospitalized
bipolar patients found an identical mean length of 7
months for manic and depressive episodes.
Wertham (1929) provided the most conclusive data on
episode length with his investigation of 1000 male and
1000 female first admissions for mania. From his published histograms we can today estimate the length of
episodes as lognormally distributed, with a mode between
2– 4 months and a median duration of 4 – 6 months (Q 1 5
2– 4 months, Q 3 5 8 –10 months). There was no
follow-up; most of these manic cases may therefore have
been bipolars.
Rennie (1942) found that there was a lengthening of
repeated manic episodes, with the first episode lasting 3.5
months, the second 5.2 months, the third 5.8 months, and
the fourth and fifth even longer. Rennie ascribed this to a
prolonging effect of aging, since before the age of 45
episode length remained constant.
Unlike Rennie, Kinkelin (1954), in a longitudinal study,
found no systematic change in episode length between the
first and seventh episodes. Among 347 bipolar patients the
mean length of depressive episodes varied impressively
from 3.5 to 8.4 months, and of manic episodes from 4 to
11.6 months. Kinkelin, too, concluded that later episodes
tended to be longer.
In the NIMH Collaborative Study on the Psychobiology
of Depression Clinical Studies (Keller et al 1986a) sur-
vival analysis demonstrated that mixed or cycling episodes
were slower to recover than pure depressed or pure manic
episodes. Perugi et al (1997) also reported that episodes
involving mixed states lasted significantly longer (mean
13.4 months) than manic episodes (8.8 months).
The computation of the length of episodes has to take
into account their lognormal distribution (Angst and Weis
1967) and to control for the individual number of episodes
(Slater 1938). In patients experiencing multiple cyclic
(bipolar) episodes, the episodes tend to be slightly shorter,
whereas in general the latest episode tends to be longer
because of the development of chronicity in some cases, as
found in our multicenter study on the course of mood
disorders (Angst et al 1973).
In the Baltimore Epidemiologic Catchment Area Follow-Up Eaton et al (1997) found a median length of
episodes of approximately 8 to 12 weeks; this is shorter
than that reported in treated populations, but similar to the
median episode length of 8 weeks reported in a community study of adolescents in Oregon (Lewinsohn et al
1994).
The most recent data from the Zurich follow-up study
(Angst and Preisig 1995a) showed a mean episode length
of 4.3 months (s 5 5.44) calculated on the basis of
intraindividual means. The median length of episodes is
not obviously dependent on the total number of episodes—see Table 2, where subgroups with two to 10
episodes are computed separately (following Slater’s
[1938] suggestion). On the whole, the median length of
episodes in bipolar illness (on the basis of individual
medians) was 3 months (Q 1 5 2 months, Q 3 5 5
months). There is a difference in length dependent on
psychopathology. Pure manic and pure depressive episodes lasted 3 months (Q 1 5 2, Q 3 5 5), as did mixed
episodes (M 5 3; Q 1 5 2, Q 3 5 7); in contrast, cyclic
episodes lasted almost 50% longer (M 5 4.19; Q 1 5
2.5, Q 3 5 7.75). These figures are based on intraindividual medians. We found no gender differences in median episode length.
The shorter episode length reported by modern studies
may be a result of their including milder cases or, what is
more probable, a consequence of antidepressant therapy:
medication has usually to be maintained for a further 6
months after recovery to avoid relapse into the still
persistent latent episode, which represents the natural
history.
Recurrence of Bipolar Disorder
The total number of episodes experienced by patients is
amply reported. Findings regarding the occurrence of
single episode cases vary widely, ranging from 0% to 55%
Bipolar Disorder: Natural History
BIOL PSYCHIATRY
2000;48:445– 457
449
Table 2. Total Number of Episodes and Length (Zurich Study)
Episodes
2
3
4
5
6
7
8
9
10
Episodes
Patients
4
8
9
15
14
21
12
13
15
Median length of episodes (months)
1.5
6.0
3.0
5.5
2.5
3.0
2.9
3.5
4.5
5.8
2.8
5.0
4.5
5.5
5.0
3.0
4.5
3.5
9.0
6.0
4.6
5.0
4.0
2.0
4.0
3.0
9.0
4.0
3.8
4.3
3.0
4.0
3.5
5.8
2.8
3.5
2.4
3.0
4.5
2.9
3.0
2.5
2.5
3.0
4.0
4.0
2.5
3.0
4.0
2.0
3.0
1.0
3.0
1
2
3
4
5
6
7
8
9
(Goodwin and Jamison 1990), with five of 11 studies
reporting rates between 0% and 8%. It is clear that any loss
of information will inflate nonrecurrence rates; furthermore, the length of follow-up and development into
chronicity have to be taken into account in assessing
findings.
The earlier literature assumed that periodic mania had a
better prognosis with fewer episodes and better outcome
than periodic melancholia or circular disorder, the latter
having the worst outcome (Rehm 1919, 107); however,
Rehm also found twice as many manic cases (53%) than
circular cases (26%) with short free intervals between
episodes (up to 1 year), reflecting a higher periodicity of
mania (it is unclear whether the length of observation in
these cases was the same).
Certain methodological advances contributed to the
investigation of the question of recurrence. One was the
introduction of a more precise terminology and clear
definitions of remission, recovery, relapse, and recurrence
(Frank et al 1991). Another was the introduction of life
table analyses into psychiatry, a method first used for
length of hospitalization (Kramer 1969) and for follow-up
data after recovery from depression (Klerman et al 1974)
and soon also applied in longitudinal studies of bipolar
disorder (Dunner et al 1976, 1979; Fleiss et al 1976, 1978).
Lavori et al (1984) applied life table methods to reanalyze
40 earlier studies on the relapse/recurrence of affective
disorders; the results varied considerably between the
studies, and the authors formulated the hypothesis of the
heterogeneity of patients’ courses in terms of low or high
hazard with a low or high risk of relapse. More recently,
survival analysis was applied to prospective data on
recurrence after remission (Fleiss et al 1978; Gitlin et al
1995; Keller et al 1993; Lavori et al 1984).
In a 2-year placebo-controlled lithium study Fleiss et al
(1978) found that under placebo 80% of bipolar I patients
experienced recurrences within about 70 weeks. (Admittedly a treatment study of this type dealt with a selected
sample of patients, who, for instance, had had a mean
6.0
10
number of two episodes in the previous 2 years.) The
4-year follow-up of a naturalistic study of mania showed
recurrence in 72% of patients, with a mere 28% remaining
in remission (Tohen et al 1990). Patients who were
symptomatic at the 6-month follow-up had a 45% greater
chance of a recurrence within the next 3.5 years.
Course of Subtypes of Bipolar Disorder
Important findings from a naturalistic study, the National
Institute of Mental Health (NIMH) Collaborative Program
on the Psychobiology of Depression, Clinical Studies
(Keller et al 1993), showed a high rate of recurrence for
pure mania (48% by 1 year and 81% by 5 years) and even
higher rates for the mixed cycling group (57% by 1 year
and 91% by 5 years). Over 7 years the rate of recurrence
was 81%. The length of sustained recovery was associated
with a lower risk for recurrence over the subsequent 4
years, but over a period of 10 years this predictive power
decreased considerably (Coryell et al 1995b); the authors
showed that even under sustained lithium prophylaxis
recurrences were present in more than 70% of cases within
5 years of recovery. This finding is consistent with the
outpatient study of Gitlin et al (1995), in which 73% of 82
bipolar patients had relapses/recurrences over an average
of 4.3 years despite maintenance pharmacotherapy (two
thirds of patients who relapsed experienced multiple relapses). Among the 26 patients who suffered no relapse
46% continued to show significant symptoms of mania or
depression.
The NIMH Collaborative Program on the Psychobiology of Depression, Clinical Studies (Coryell et al 1989)
provided no evidence that the course of bipolar II disorder
differed from that of bipolar I, which confirms our own
findings (Angst 1986).
A recent study comparing patients suffering from bipolar II disorder with major depressive subjects under
fluoxetine showed similar relapse/recurrence rates of 36%
versus 35% after 50 weeks and 44% versus 49% after 62
450
BIOL PSYCHIATRY
2000;48:445– 457
weeks (Amsterdam et al 1998). This study provides no
indication of a deterioration in the course of bipolar II
disorder due to antidepressant medication; on the contrary,
the disorder had a better course than under placebos.
Marneros et al (1991a) reported that bipolar disorder
and schizo-bipolar disorder had very similar course
characteristics.
Is Recurrence of Bipolar Illness
Progressive?
Describing the intervals between episodes, Kraepelin
(1913, 1365) found a progressive shortening of the first
three intervals (first interval of male/female subjects 4.6/
4.3 years, second interval 2.8/2.0 years, third interval
1.2/1.4 years). His data, like that of some other authors, are
not controlled for the total number of episodes: clearly the
intervals between episodes in patients having few episodes
over a lifetime will be longer than in patients who
experience many. We owe this particular methodological
breakthrough to Eliot Slater (1938), who, in his paper on
the periodicity of manic– depressive insanity, investigated
116 patients of the Forschungsanstalt für Psychiatrie in
Munich who had been personally diagnosed by Kraepelin.
Studying the length of intervals between episodes, Slater
made his major methodological contribution by controlling for the total number of episodes and analyzing
separately the subgroups with 1, 2, 3, . . . , n intervals. He
showed that there was indeed a shortening of the intervals,
but only between the first few manifestations of the illness.
Investigating the individual periodicity, he concluded that
every patient had his or her own rhythm.
In 1968 Bratfos and Haug applied Slater’s method to the
follow-up data on 215 cases of manic– depressive disorder
(including depression) in an analysis of the length of the
intervals between episodes. Correcting for the number of
episodes, the authors found that the first interval length
was 2.1 years, the second was 4.8 years, and the third was
2.2 years; thus no clear tendency emerged from these data.
Again following Slater’s method of correcting for the
number of episodes, Angst et al (1973) published the
results of a multinational retrospective and partially prospective hospital record study on the course of 393 bipolar
and 634 unipolar depressive patients. Analyzing cycle
lengths, they indirectly confirmed Slater’s finding that the
intervals shortened as the number of recurrences increased; however, the median cycle lengths gave clear
evidence only for continuous shortening of the first three
cycles; the pattern of later recurrences seemed to be
unpredictable. On the other hand, nonparametric tests
between successive cycle lengths showed a significant
shortening from cycle one to cycle 11 (Angst et al 1973,
499). Marneros et al (1991a) confirmed the shortening of
J. Angst and R. Sellaro
cycles on the basis of the Cologne naturalistic follow-up
study of 30 bipolar and 56 schizo-bipolar patients. Zis et al
(1980) and Zis and Goodwin (1979) arrived at similar
conclusions. This seemed to confirm earlier findings that
had suggested decreasing cycle length (e.g., those of
Kraepelin [1913, 1325] or Kinkelin [1954], although not
controlled for number of cycles—see Figure 6-3 in Goodwin and Jamison [1990]). All these results, together with
the finding that precipitation rates decrease (Angst 1966,
41) with increasing recurrence, led Roy-Byrne et al (1985)
to speak of sensitization and Post et al (Post 1992; Post et
al 1984, 1986) to develop the theory that vulnerability
grows with the number of episodes and the theory of
conditioning, sensitization, and kindling (by analogy with
electrophysiologic kindling).
National data on hospital admissions and readmissions
can also provide a rough estimate of the natural history of
severe cases. Kessing et al (1998a, 1998b) recently described the course of Danish hospital admissions on the
basis of a nationwide register of ICD-8 diagnoses. Rehospitalizations were taken as a measure of recurrence. The
authors found, on the basis of 2903 bipolar cases, a
progressive shortening of the interval between discharge
from hospital and the next rehospitalization and, therefore,
a deteriorating course. Selection bias did not completely
explain the shortening intervals between hospital admissions, and control for gender and age did not alter the
conclusions. So far, then, it would seem to have been
established that the course of bipolar disorder is recurrent
and progressive, but this aspect is still surrounded by
considerable controversy, which may partially be due to
the possibility “that recent studies deal with different,
more broadly diagnosed populations than the seminal,
earlier studies” (Grof et al 1995).
The Lithium Controversy: Does Recurrence
Improve Spontaneously?
One controversy about the natural history of bipolar
disorder dates back to 1968 and the criticism by two
reputed British authors, Blackwell and Shepherd of lithium trials. They assumed not only the inefficacy of lithium
but also that bipolar disorder had a good prognosis,
making long-term prophylaxis unnecessary. This view was
supported theoretically by Lader (1968) and empirically
by Saran (1969), who found no evidence for high recurrence in his follow-up data and who concluded that past
recurrence was not predictive for recurrence in the future.
Saran’s findings were of critical importance because the
early work on lithium had been based statistically on the
assumption that high recurrence in a patient’s previous
history should be expected to repeat itself in the future.
Saran concluded that his findings on the spontaneous
Bipolar Disorder: Natural History
course corresponded to the course as observed under
lithium treatment by Baastrup and Schou (1968). Saran’s
findings have not been confirmed by other studies: a
methodological investigation by Isaaksson et al (1969)
demonstrated the persistence of recurrence for bipolar and
unipolar depression, and the same conclusion was reached
by Laurell and Ottosson (1968).
Stability of Recurrence after Initial
Deterioration
Another debated question is whether the course of bipolar
disorder is really progressive, characterized by unlimited
shortening of cycles throughout (Angst et al 1973). This
model is disconfirmed by a number of new studies. The
recent summary of the literature by Kessing et al (1998b)
shows that Fukuda et al (1983), who investigated not the
early but the later course of the illness, could find no
shortening of cycles (a finding that would still be compatible with the hypothesis that shortening is a feature of the
first few cycles only). This aspect of the systematic
shortening of cycle length was also very seriously questioned in the reviews by Coryell and Winokur (1992) and
Solomon et al (1995), which drew mainly on Winokur and
colleagues’ findings (1994) from the prospective naturalistic NIMH study covering 10 years (Turvey et al 1999).
This study found that the second cycle was clearly shorter
than the first, the third a little longer than the second, but
the fourth and fifth cycles were again shorter. On the other
hand, Winokur et al (1993) stressed that bipolar illness
was highly recurrent, with an “inexorable continuation of
episodes and hospitalisations,” and could find no data
suggesting that the illness burned out at a later stage.
On the basis of prospective data Coryell and Winokur
(1992) found that rapid cycling, which is observed in 20%
to 25% of bipolar patients, is usually a transient manifestation and not therefore a characteristic of the long-term
course.
Meta-Analysis of Two Studies on the Course
Samples and Methodology
Here we reanalyze some data from our two studies, the
mainly retrospective multicenter study published by Angst
et al (1973) and the Zurich follow-up study described by
Angst and Preisig (1995a).
1. The multicenter study (Angst et al 1968a, 1973)
consisted of consecutive hospital admissions of
bipolar and unipolar depressive patients from Basel
(103); Berlin (104); Hamilton, Canada (69); Prague
(132); Zurich (392); Glostrup, Denmark (100); and
Landeck, Germany (140). For the present analysis
BIOL PSYCHIATRY
2000;48:445– 457
451
the Zurich sample was excluded, to avoid any
overlap with the Zurich follow-up study. The remaining sample consisted of 329 bipolar patients.
The data collection was mainly of the retrospective
type based on case histories and verbal information.
The documentation of the course was carried out by
means of a standardized form of protocol in which
all data were entered separately for each episode and
subsequent interval (Angst and Weis 1968). In
retrospect the onset (date) and length (months,
weeks) of previous episodes and aspects of treatment (none, ambulatory, hospital) were assessed.
The degree of remission was coded as full, partial,
or unknown. Psychopathology was coded with a list
of 10 syndromes. The data were reanalyzed for this
study.
2. The Zurich follow-up study consisted of 406 consecutive hospital admissions for severe depression
or mania from 1959 to 1963. Regular follow-up
investigations (by telephone, interviews, and record
collection) were conducted in 1963, 1965, 1979,
1975, 1980, and 1985. In 1991 and 1997 mortality
data were available from the Swiss federal office of
statistics. Seventy-six percent of patients had died
by the end of 1997. The sample and the methodology of assessments were described in detail by
Angst and Preisig (1995a). The principal data collected were comparable to the multicenter study, but
psychopathology and treatment were assessed in
more detail. The 220 bipolar patients were reanalyzed for this article.
Results
Table 3 presents the data from the mainly retrospective
multicenter study and demonstrates a systematic shortening of the first four cycles. The predominantly prospective
data from the Zurich sample (Table 4) shows a significant
shortening between cycles 1 and 2 only, with no systematic change thereafter. In both studies the conclusions were
confirmed by t tests for dependent measures.
In a survival analysis of the Zurich follow-up data
significant differences were found between cycles 1–5 but
were difficult to interpret. The first cycle was longer and
the second cycle shorter than all the others. Otherwise the
survival curves were very similar (Figure 3). The mean
cycle length was 28.7 months (s 5 30.93) but the
median length, which is much more representative, only
18 months (Q 1 5 3, Q 3 5 18); exclusion of the first
longer cycle does not change this median.
In conclusion, we found a shortening of cycle length at
the beginning of the disorder only; later episodes were
persistently recurrent but occurred at irregular intervals
without any systematic deterioration or amelioration, thus
452
J. Angst and R. Sellaro
BIOL PSYCHIATRY
2000;48:445– 457
Table 3. Total Number of Cycles and Length (Multicenter Study)
Cycles
1
2
3
4
5
6
7
8
9
Patients
18
35
38
45
39
31
26
27
17
Cycles
Median length of cycles (months)
23
26
38
21
35
42
28
42
37
15
20
27
18
31
18
17
20
15
13
19
19
16
14
18
14
13
15
12
14
9
11
15
13
9
8
9
15
11
12
12
9
15
10
15
12
1
2
3
4
5
6
7
8
9
This table does not include Zurich data.
confirming Winokur and colleagues’ (1993, 1994)
findings.
Episode Frequency per Year
Periodicity can also be expressed by the episode frequency
per year. In a follow-up of 140 bipolar I patients over 11.4
years Dunner et al (1979) found 0.54 episodes per year; we
found 0.37 (Angst and Preisig 1995a). Marneros’ group
found 0.41 episodes per year (Marneros 1999; Marneros et
al 1991a) for bipolar disorder and about half as many in
the case of recurrent depression. These figures included
the first cycle, which is considerably longer than later
ones. Dunner et al (1979) found no relationship between
episode frequency and age of onset (which was not
consistent with our early results; Angst and Weis 1967).
The question of a gender difference in episode frequency
per year remained open. After the onset of their disorder,
bipolar patients spent on the average about 19% of their
lives in affective episodes over an observation period of 27
years (Angst and Preisig 1995a).
In the present analysis the median cycle length (on the
basis of individual medians) was 18 months (Q 1 5 12
months, Q 3 5 33.5 months), a length that corresponds to
a recurrence rate of 0.66 episodes per year. On the basis of
an episode length of 3 months (median), bipolar patients
spent about 2 months/year in episodes. We could not find
any gender differences (Mann–Whitney U test) in episode
frequency measured by cycle length. Recently Gottschalk
et al (1995) found some evidence for a nonlinear deterministic structure in long-term daily mood records of six
out of seven bipolar patients, something not found in
normal control subjects. Further research on a chaotic
course is certainly desirable.
Correlates of Recurrence
For the analysis of correlates with long-term recurrence
we excluded the first two cycles and entered the median of
all intraindividual medians of cycle length as the dependent variable into a multiple regression analysis. Cycle
length did not correlate with gender, retrospective versus
prospective data collection, or long-term administration of
tricyclic antidepressants. This finding does not confirm the
assumed deterioration of the course of bipolar illness
under tricyclics (Arnold and Kryspin-Exner 1965; Koukopoulos et al 1980; Till and Vuckovik 1970; Tondo et al
1981). Unsurprisingly, long-term lithium and antipsy-
Table 4. Total Number of Cycles and Length (Zurich Study)
Cycles
1
2
3
4
5
6
7
8
91
Cycles
Patients
4
8
9
15
14
21
12
13
124
Median length of cycles (months)
64
105
55
78
30
44
42
34
26
12
21
25
22
35
23
22
19
34
33
37
36
34
23
21
45
19
24
16
30
15
20
24
17
23
15
17
30
12
12
17
17
12
28
12
1
2
3
4
5
6
7
8
12
91
Bipolar Disorder: Natural History
BIOL PSYCHIATRY
2000;48:445– 457
453
Figure 3. Survival analysis of cycle lengths (first to
fifth).
chotic medication correlated with shorter cycles; this
result is explicable by the selection of highly recurrent
cases for prophylaxis. The syndromal characteristics of
episodes had some impact on cycle length: it was significantly shorter in cases with cyclic versus pure (manic or
depressive) episodes and in those with more manic than
depressive episodes.
Poor Outcome of Bipolar Disorder in Early
Follow-Up Studies
In the predrug era most bipolar cases were described as
manifesting residual symptoms after recovery from episodes. Their state was described as “unsteady, moody,
irritable, indolent, egocentric” (Pilcz 1901, 61– 62), a
symptom that often preceded the episodes or had even
been present since childhood. Kraepelin (1913, 1185), too,
although assuming a good prognosis, admitted the existence of mild residual “debility states.” Ziehen (1902)
estimated complete recovery in periodic mania to occur at
most in 20% of cases.
In a follow-up of first admissions from 1920 to 1947
over a mean of 22 years Kinkelin (1954) found chronicity
(including severe residual states) in 14.6% of depressive
subjects (N 5 89) and in 41% of circular cases (N 5
51).
An important follow-up study of 297 bipolar and 945
unipolar patients in the Phipps Clinic (admissions from
1913 to 1940) was published by Stephens and McHugh
(1991). Compared with depressive subjects, bipolar patients had an earlier age of onset on admission and were
more likely to be psychotic; they also had more serious
premorbid characteristics, more sudden onsets (44%),
more previous admissions (62% for depression and 67%
for mania), and more problems with alcohol, but more
bipolar than depressed patients were discharged in a
recovered state. In bipolar disorders they found a lifetime
average of 3.6 hospitalizations, and in their follow-up they
observed 0.19 episodes of depression and 0.29 episodes of
mania per year followed up. Only 2% of all patients had
manic episodes with no lifetime depressions; this small
group had the best outcome. Bipolar patients (N 5 301)
had the poorest outcome: 7% were recovered, 50% improved, and 43% unimproved at the follow-up rating. The
rates of recovered, improved, and unimproved patients
were 19%, 56%, and 25% for pure mania (N 5 16) and
25%, 44%, and 32% for pure depression (N 5 700).
In a follow-up of 86 manic patients over 30 to 40 years,
outcome (measured by psychiatric symptoms) was good in
43% of cases, fair in 18%, and poor in 25% (Tsuang et al
1979).
Poor Outcome in Recent Follow-Up Studies
Modern studies (e.g., a 4.5-year follow-up study by
Goldberg et al [1995]) showed that only 41% of bipolar
patients had a good overall outcome, the remainder being
moderately impaired (37%) or showing poor functioning
(22%). Gitlin et al (1995) found a poor outcome even
under prophylactic medication (mainly lithium); poor
outcome was more closely associated with the number of
depressive episodes than the number of manic episodes. In
their above-mentioned NIMH Collaborative Program on
the Psychobiology of Depression with a 15-year intensive
follow-up, Coryell et al (1998) found that 56.6% of 113
bipolar I patients had had no symptoms over the past 12
months, 22.1% had had symptoms for fewer than 52
weeks, and 20.4% had had symptoms in all 52 weeks,
representing a poor outcome (defined as having had
symptoms of major depressive disorder, mania, or schizoaffective disorder in all 52 weeks of year 15).
The short-term outcome, 12 months after discharge
from hospital, was similar in patients with manic episodes
or mixed episodes (Keck et al 1998), whereas in Keller
and colleagues’ (1986b) study after 18 months of follow-up mixed episodes also developed more frequently
454
J. Angst and R. Sellaro
BIOL PSYCHIATRY
2000;48:445– 457
Table 5. Outcome of 219 Patients in the Zurich Follow-Up
Study (Median Age at Follow-Up or Death, 68 Years)
Recovered (GAS score . 60, no episodes over the past
5 years)
Remitted (GAS score . 60) but still recurrent (,5
years since last episode)
Incomplete remission (GAS score 1– 60) over more
than 5 years
Incomplete remission, course still recurrent
Chronic (last episode without remission, minimum
length 2 years)
Suicide
16.0%
25.5%
2.
7.8%
27.0%
15.9%
3.
7.8%
GAS, Global Assessment Scale.
into chronicity (32%) than pure manic episodes (7%). In
bipolar I patients poor outcome did not correlate clearly
with early onset of the disorder or cycling per se (Coryell
et al 1998), but persistence of depressive symptoms in
years 1 and 2 was correlated with impairment after 15
years (household duties, recreational activities, overall
satisfaction, and global social adjustment). Such early
persistence of depressive symptoms predicted a poor
prognosis of a bipolar subtype, whereas this was not the
case for early persistence of manic symptoms.
The lifetime outcome of bipolar disorder in our Zurich
follow-up study is given in Table 5 and demonstrates a
poor prognosis despite modern treatments. Up to a median
age of 68 years only 16% of patients had recovered; 52%
still suffered from recurrent episodes and the remaining
patients had become chronically ill or had committed
suicide. These data underline the poor outcome into old
age and the need for intensive treatment. This table shows
the outcome of bipolar disorder before the occurrence of
an organic brain syndrome in the elderly, which was found
in 14.5% of cases. The previous number of affective
episodes was not correlated with the development of an
organic brain syndrome (Angst and Preisig 1995b).
Future studies on outcome should clearly distinguish
between different outcome measures, since Tohen et al
(2000) have shown that functional recovery can be much
worse than syndromal recovery.
Conclusions
Several important conclusions regarding episodes, recurrence, and outcome have emerged from this review and
data analysis of the natural history of bipolar disorder:
1. Before the introduction of modern drugs, spontaneous mild depression following a manic episode and
spontaneous hypomania following a melancholic
episode were very common; they were interpreted as
“reactive” and had no effect on the principal diagnosis of pure mania or pure melancholia. This
4.
5.
6.
7.
historical fact has to be taken into account in today’s
hypothesis of drug-induced hypomania or druginduced depression; these have to be proven statistically by placebo-controlled trials.
Our decades-long prospective study shows that over
lifetime the proportions of mania and depression in
bipolar disorder remain stable into old age. Bipolar
female subjects manifest more depression than bipolar male subjects.
Mixed states have been described since the early
19th century. Modern studies have demonstrated
that they have a poorer prognosis than other bipolar
conditions, with slower remissions and higher risk
for chronicity.
The natural length of affective episodes has probably not changed over the past 120 years. Patients
responding to antidepressants still require a maintenance treatment throughout the underlying episode.
In clinical studies the median length of episodes is 3
to 6 months; in epidemiological studies it is 2 to 3
months.
The recurrence of bipolar disorder was always the
rule; it now seems to be established that there is
some initial shortening of intervals/cycles, followed
by an irregular persistent recurrence, with a median
cycling of 18 months. In contrast to earlier reports,
the new studies show that there is no unlimited
shortening of cycle length—not therefore supporting
the kindling model. In several studies rapid cycling
has been found to be relatively frequent, but usually
transient.
Lifelong outcome has rarely been studied, and
precise data on the natural outcome are scarce. Some
new prospective studies demonstrate that most patients continue to suffer from residual depressive or
hypomanic symptoms between episodes, and many
are functionally impaired.
Overall research into the natural history of bipolar
illness shows that it has a poor prognosis, as
reflected by high recurrence, chronicity of episodes
or residual symptoms, and premature death by
suicide and somatic disorders; however, unlike
schizophrenia, which is characterized by much
higher chronicity and the predominance of negative
and psychotic symptoms, chronicity in bipolar disorders is rarer (10 –20%) and the more frequent
residual states are limited to characteristic depressive and hypomanic symptoms.
The authors thank Professor Andreas Marneros (University of Halle) and
Dr. Mauricio Tohen (Lilly Company) for their contribution to this article
in the form of comments and suggestions.
Bipolar Disorder: Natural History
Aspects of this work were presented at the conference “Bipolar
Disorder: From Pre-Clinical to Clinical, Facing the New Millennium,”
January 19 –21, 2000, Scottsdale, Arizona. The conference was sponsored by the Society of Biological Psychiatry through an unrestricted
educational grant provided by Eli Lilly and Company.
References
Amsterdam JD, Garcia-Espana F, Fawcett J, Quitkin FM, Reimherr FW, Rosenbaum JF, et al (1998): Efficacy and safety of
fluoxetine in treating bipolar II major depressive episode.
J Clin Psychopharmacol 18:435– 440.
Angst J (1966): Zur Aetiologie und Nosologie Endogener Depressiver Psychosen. Eine Genetische, Soziologische und
Klinische Studie. Berlin: Springer.
Angst J (1978): The course of affective disorders. II. Typology of
bipolar manic-depressive illness. Arch Psychiatr Nervenkr
226:65–73.
Angst J (1986): The course of major depression, atypical bipolar
disorder, and bipolar disorder. In: Hippius H, Klerman GL,
Matussek N, editors. New Results in Depression Research.
Berlin: Springer, 26 –35.
Angst J (1987): Switch from depression to mania, or from mania
to depression. J Psychopharmacol 1:13–19.
Angst J (1988): Clinical course of affective disorders. In:
Helgason T, Daly RJ, editors. Depressive Illness: Prediction
of Course and Outcome. Berlin: Springer, 1– 48.
Angst J, Baastrup PC, Grof P, Hippius H, Poeldinger W, Weis P
(1973): The course of monopolar depression and bipolar
psychoses. Psychiatr Neurol Neurochir 76:489 –500.
Angst J, Battegay R, Bente D, Berner P, Broeren W, Cornu F, et
al (1968a): Das Dokumentationssystem der Arbeitsgemeinschaft für Methodik und Dokumentation in der Psychiatrie
(AMP). Arzneimittelforschung 18:3– 8.
Angst J, Grof P, Hippius H, Poeldinger W, Weis P (1968b): La
psychose maniaco-depressive est-elle periodique ou intermittente? Dans: Cycles biologiques et psychiatrie. In: de Ajuriaguerra J, editor. Symposium Bel-Air III, Geneve 1967.
Geneva: Georg et Cie, 339 –351.
Angst J, Preisig M (1995a): Course of a clinical cohort of
unipolar, bipolar and schizoaffective patients. Results of a
prospective study from 1959 to 1985. Schweiz Arch Neurol
Psychiatr 146:5–16.
Angst J, Preisig M (1995b): Outcome of a clinical cohort of
unipolar, bipolar and schizoaffective patients. Results of a
prospective study from 1959 to 1985. Schweiz Arch Neurol
Psychiatr 146:17–23.
Angst J, Weis P (1967): Periodicity of depressive psychoses. In:
Brill H, Cole JO, Deniker P, Hippius H, Bradley PB, editors.
Neuro-Psychopharmacology. Proceedings of the Fifth International Congress of the Collegium Internationale NeuroPsychopharmacologicum, Washington, D.C. 1966, International Congress Series No 129. Amsterdam: Excerpta
Medica, 703–710.
Angst J, Weis P (1968): Aetiologie und Verlauf endogener
Depressionen. In: Rémy M, editor. Depressionen und ihre
Behandlung. Fortbildungskurse der SGP, Vol 1. Basel, Switzerland: Karger, 8 –16.
Arnold OH, Kryspin-Exner K (1965): Zur Frage der Beeinflus-
BIOL PSYCHIATRY
2000;48:445– 457
455
sung des Verlaufes des manisch-depressiven Krankheitsgeschehens durch Antidepressiva. Wien Med Wochenschr 115:
929 –934.
Baastrup PC, Schou M (1968): Prophylactic lithium. Lancet
I:1419 –1422.
Baillarger J (1854): De la folie à double forme. Ann Med Psychol
6:369 –384.
Ballet G (1903): Traité de Pathologie Mentale. Paris: Doin.
Blackwell B, Shepherd M (1968): Prophylactic lithium: Another
therapeutic myth? An examination of the evidence to date.
Lancet 1(7549):968 –971.
Bourgeois ML, Marneros A (in press): Course and prognosis of
bipolar disorders. In: Marneros A, Angst J, editors. Bipolar
Affective Disorders. Dordrecht, The Netherlands: Kluwer.
Bratfos O, Haug JO (1968): Course of manic-depressive psychosis. A follow-up investigation of 215 patients. Acta Psychiatr
Scand 44:89 –112.
Coryell W, Endicott J, Maser JD, Keller MB, Leon AC, Akiskal
HS (1995a): Long-term stability of polarity disinctions in the
affective disorders. Am J Psychiatry 152:385–390.
Coryell W, Endicott J, Maser JD, Mueller T, Lavori P, Keller M
(1995b): The likelihood of recurrence in bipolar affective
disorder: The importance of episode recency. J Affect Disord
33:201–206.
Coryell W, Keller M, Endicott J, Andreasen N, Clayton P,
Hirschfeld R (1989): Bipolar II illness: Course and outcome
over a five-year period. Psychol Med 19:129 –141.
Coryell W, Turvey C, Endicott J, Leon AC, Mueller T, Solomon
D, Keller M (1998): Bipolar I affective disorder: Predictors of
outcome after 15 years. J Affect Disord 50:109 –116.
Coryell W, Winokur G (1992): Course and outcome. In: Paykel
ES, editor. Handbook of Affective Disorders, 2nd ed. London:
Churchill Livingstone, 89 –108.
Dunner DL, Fleiss JL, Fieve RR (1976): The course of development of mania in patients with recurrent depression. Am J
Psychiatry 133:905–908.
Dunner DL, Murphy D, Stallone F, Fieve RR (1979): Episode
frequency prior to lithium treatment in bipolar manic-depressive patients. Compr Psychiatry 20:511–515.
Eaton WW, Anthony JC, Gallo J, Cai G, Tien A, Romanoski A,
et al (1997): Natural history of Diagnostic Interview Schedule
DSM-IV major depression. The Baltimore Epidemiologic
Catchment area follow-up. Arch Gen Psychiatry 54:993–999.
Falret J (1861): Principes à suivre dans la classification des
maladies mentales. Ann Med Psychol 19:145.
Falret JP (1851): Marche de la folie. Gazette Hopitaux 24:18 –19.
Falret JP (1854): De la folie circulaire ou forme de maladie
mentale caractérisée par l’alternative régulière de la manie et
de la mélancholie. Bull Acad Med (Paris) 19:382.
Fleiss JL, Dunner DL, Stallone F, Fieve RR (1976): The life
table: A method for analyzing longitudinal studies. Arch Gen
Psychiatry 33:107–112.
Fle
Historical Perspectives and Natural History of Bipolar
Disorder
Jules Angst and Robert Sellaro
A review of two centuries’ literature on the natural history
of bipolar disorder, including modern naturalistic studies
and new data from a lifelong follow-up study of 220
bipolar patients, reaches the following conclusions: the
findings of modern follow-up studies are closely compatible with those of studies conducted before the introduction of modern antidepressant and mood-stabilizing treatments. Bipolar disorder has always been highly recurrent
and considered to have a poor prognosis.
Bipolar patients who have been hospitalized spend about
20% of their lifetime from the onset of their disorder in
episodes. Fifty percent of bipolar episodes last between 2
and 7 months (median 3 months). The intervals between
the first few episodes tend to shorten; later the episodes
return at an irregular rhythm of about 0.4 episodes per
year with high interindividual variability. Switches from
mania into mild depression and from depression into
hypomania were frequently reported in the 19th century
and the first half of the 20th.
Antidepressant and antimanic drugs have to be given as
long as the natural episode lasts. Given the poor outcome
of bipolar disorders found in naturalistic follow-up studies
and our lifelong investigation, intensive antidepressant,
antimanic, and mood-stabilizing treatments are required
in most cases. Despite modern treatments the outcome into
old age is still poor, full recovery without further episodes
rare, recurrence of episodes with incomplete remission the
rule, and the development of chronicity and suicide still
frequent. Biol Psychiatry 2000;48:445– 457 © 2000 Society of Biological Psychiatry
Key Words: Bipolar disorder, natural history, course,
recurrence, outcome
Introduction
T
his article briefly reviews the natural history of bipolar
disorder, giving special weight to historical studies
before the era of antidepressants; integrates the results of
modern naturalistic follow-up studies; and from our own
findings 1) reanalyzes data from an early multicenter study
(Angst et al 1968b, 1973) and 2) includes some new data
From the Zurich University Psychiatric Hospital, Zurich, Switzerland.
Address reprint requests to Jules Angst, M.D., Zurich University Psychiatric
Hospital, Box 68, Lenggstreet 31, Zurich 8029, Switzerland.
Received January 13, 2000; revised April 13, 2000; accepted April 20, 2000.
© 2000 Society of Biological Psychiatry
from our lifelong Zurich follow-up study (Angst and
Preisig 1995a, 1995b). The article focuses mainly on
episodes and recurrence and, to a lesser extent, outcome; it
does not deal with rapid cycling and seasonal depression.
Recent reviews of the course of bipolar disorder have been
published by Lavori et al (1984), Keller (1987), Goodwin
and Jamison (1990), Coryell and Winokur (1992), Verdoux and Bourgeois (1995), Kessing et al (1998), Goldberg and Harrow (1999), Marneros (1999), and Bourgeois
and Marneros (in press).
The Concept of Bipolar Disorder
We owe the categorization of bipolar disorder as an illness
to Falret, who in 1851 and 1854 on the basis of longitudinal observations developed the entity of “folie circulaire” (circular madness), defined by manic and melancholic episodes separated by symptom-free intervals. In
1854 Baillarger used the term folie à double forme to
describe cyclic (manic–melancholic) episodes (Pichot
1995; Ritti 1879). Kraepelin called such cyclic episodes
“double attacks.” In both French diagnoses the prognosis
was considered to be “desperate, terrible and incurable”
(Bourgeois and Marneros, in press). Circular illness was
described by most authors as a recurrent condition; it
became the prototype of the larger group of periodic
psychoses embracing periodic mania, periodic melancholia, and periodic cyclic disorders (Ballet 1903; Mendel
1881; Pilcz 1901; Ziehen 1902, 1907).
The Concept of Mixed States
The history of the concept of mixed states has been
extensively studied by Marneros (in press): what we today
call “mixed states” were probably already known at the
beginning of the 19th century and named “mixtures”
(Mischungen) by Heinroth in 1818 and “middle forms”
(Mittelformen) by Griesinger (1845). Guislain (1852) gave
clear descriptions of different syndromes of mixed states.
The history of bipolar disorder by Haustgen (1995) traces
the term mixed states to J.P. Falret’s son Jules Falret
(1861).
Very influential in this field was Weygandt (1899), who
worked with Kraepelin and whose monograph distinguished three forms of mixed states: manic stupor, agitated
0006-3223/00/$20.00
PII S0006-3223(00)00909-4
446
BIOL PSYCHIATRY
2000;48:445– 457
melancholia (depression with flight of ideas and agitation),
and unproductive mania (elated mood, increased motor
activity, and inhibition of thinking). Kraepelin’s (1899)
textbook descriptions of mixed states were founded on
Weygandt’s monograph. Further progress was made by
Rehm’s monograph (1919, 113), which classified mixed
states systematically on the basis of the permutations of
the three elements that had been defined by Kraepelin:
thought disorder, mood, and psychomotor activity (identified as a, b, and c for mania and as A, B, and C for
depression).
Kraepelin’s Manic–Depressive Insanity
At the turn of the 19th century Kraepelin’s unifying
approach to the classification of mood disorders (1899)
resulted in bipolar disorders being subsumed within manic– depressive insanity (MDI), a broad group that included
single-episode and recurrent depression. Kraepelin (1913,
1183) was later himself to raise the possibility of the
heterogeneity of MDI. Unlike the French concepts, Kraepelin’s MDI had a good prognosis and did not develop
into severe dementia, although Kraepelin conceded the
existence of mild residual states after recovery from the
episodes themselves (Schwächezustände; Kraepelin 1913,
1349) and of mild fluctuations between episodes. Kraepelin considered periodicity to be unimportant for the diagnosis (Pilcz 1901). As a consequence of Kraepelin’s
unification of affective disorders, research on their course
frequently failed to distinguish between depression, mania, and bipolar disorder (Bratfos and Haug 1968; Fuller
1935; Paskind 1930; Pollock 1931a, 1931b, 1931c; Poort
1945; Rennie 1942; Tomasson 1947).
Notable contemporary authors nevertheless disagreed
with Kraepelin’s unitarian approach, and their studies of
the natural history of affective disorders maintained the
distinction between mania, depression, and bipolar disorder (Ballet 1903; Pilcz 1901; Ziehen 1902, 1907). This
data on the course of bipolar disorder collected in the 19th
century and the first half of the 20th, before the introduction of modern antidepressants and mood stabilizers, is of
special value in that it represents the disorder’s untreated
natural history.
Onset of Bipolar Disorder
The dating of the age of onset is to a certain extent
unreliable because it is usually retrospective and dependent on insecure recall. Bipolar disorder begins about 10
years earlier than recurrent depression, as shown by a
review of the literature (Angst 1988). Earlier studies
indicated a mean age of 28 to 33 years; epidemiologic and
newer clinical studies show that bipolar symptoms start
J. Angst and R. Sellaro
frequently in adolescence (Weissman et al 1988) and that
manic episodes manifest usually in the early 20s (Fogarty
et al 1994).
Periodic Mania and Switches of Polarity
There is considerable interest today in data on the course
of single and multiple episodes, which can answer questions about the psychopathology, duration, and frequency
of episodes; the syndromal stability over lifetime; and the
frequency with which initial major depression develops
into bipolar disorder. Today the switch of an episode from
depression to hypomania is often assumed to be drug
induced, but the phenomenon was already very common
as “reactive hyperthymia” before the introduction of
antidepressants.
A century ago the concept of periodic mania was well
known and the diagnosis much more frequent, quite
simply because it was applied to cases that today would be
considered bipolar disorders. For instance, an initial depressive syndrome cycling into a manic episode, although
frequently observed, was not considered an indication of
bipolarity (Mendel 1881; Ziehen 1902). Similarly, “postmelancholic reactive hyperthymia” with clear hypomanic
symptoms was compatible with the diagnosis of pure
periodic melancholia (Ziehen 1907, 26).
Mania switching into depression was likewise very
commonly reported as “reactive depression” (Ziehen
1902, 546, 554; Wernicke 1906, 355). Postmanic depression lasted a few days or a few weeks according to
Wernicke (1906, 355). Mild depression was observed
preceding or terminating manic attacks in most of the 128
manic patients studied by MacDonald (1918).
Modern naturalistic and treatment studies have also
found that mania frequently cycles into depression: the
rates of cycling observed in follow-up studies over 8
weeks vary from 17% (Tohen et al 1990) to 30% (Keller
et al 1986a). Our earlier retrospective record study of 300
manic patients (admitted between 1920 and 1970) found
that 21% of manic episodes cycled into depression, a rate
that did not change significantly during the intervening
decades (Angst 1987).
In a retrospective record study, depression switching
into hypomania was found in 29% of bipolar patients
hospitalised between 1920 and 1959 (Angst 1987).
Diagnostic Change from Depression to
Bipolar Illness
The syndromal course over lifetime has been little investigated. It was frequently assumed that mania was predominant in earlier years and depression in the second half of
life. Kinkelin (1954) followed-up 146 hospital first admis-
Bipolar Disorder: Natural History
BIOL PSYCHIATRY
2000;48:445– 457
447
Figure 1. Proportion of syndromes across
20 episodes in male subjects. BP, bipolar;
Man, mania; Dep, depression.
sions suffering from MDI (1929 –1947) until 1948 in a
study covering an average of 21.8 years of the total course
of the disorder. Of the 146 cases, 125 began with depression and 21 with mania. During the follow-up period 36
(28.8%) of the 125 depressive patients developed manic
episodes, a figure that would correspond to a diagnostic
change from depression to bipolar illness of 1.3% per year
of observation. Marneros et al (1991b) reported that the
initial diagnosis of depression remained stable in 79% of
cases over 27 years. In our preponderantly prospective
study we found a rate of diagnostic change from depression to hypomania/mania of about 1% per year (Angst and
Preisig 1995a). Coryell and colleagues’ (1995a) intensive
prospective follow-up study of 381 depressive subjects
over 10 years found that 10.2% developed into mania
(5.2%) and hypomania (5.0%), which also corresponds to
a 1% change per year of observation.
In their monograph, Marneros et al (1991a) found that
mania frequently developed into schizomania or mixed
states.
Syndromal Stability of the Course
New data from our Zurich follow-up study confirmed a
major gender difference in the psychopathology of bipolar
patients over the first 20 episodes (Angst 1978): female
subjects manifested significantly more depressed episodes
and male subjects more cyclic episodes (mania and depression), whereas pure manic and mixed episodes were
equally frequent in both genders (Figures 1 and 2). The
syndromal proportions were found to remain remarkably
stable over 20 episodes (Angst and Weis 1967), which also
means, for instance, that aging brings no increase in the
depressive component of bipolar illness.
Figure 2. Proportion of syndromes across
20 episodes in female subjects. BP, bipolar;
Man, mania; Dep, depression.
448
J. Angst and R. Sellaro
BIOL PSYCHIATRY
2000;48:445– 457
Table 1. Length of Episodes (Months) of Patient Samples
Mean
Mendel (1881)
Kraepelin (1913)
Panse (1924)
Wertham (1929)
Rennie (1942)
Kinkelin (1954)
Angst and Preisig (1995a)
Median
Q1
Q3
5– 6
3– 4
6 –7
4–6
2– 4
8 –10
3d
2
5
6–8
7
3.5a–5.8b
3.5– 8.4
4.3c
a
First episode.
Third episode.
Mean after logarithmic transformation.
d
Median of Table 2.
b
c
Natural Length of Episodes
Valuable data on the natural length of episodes were
published before the introduction of effective treatments
(Table 1).
Mendel (1881, 155) reported data on the length of
manic episodes (N 5 43), from which a median duration
of 5– 6 months (Q 1 5 3– 4 months, Q 3 5 6 –7 months)
can be computed; Mendel found only one episode which
lasted 10 –12 months and one over 1 year. Kraepelin
(1913) stressed the great variability in episode length but
estimated that most episodes lasted between 6 and 8
months.
Panse’s follow-up study (1924) of 205 hospitalized
bipolar patients found an identical mean length of 7
months for manic and depressive episodes.
Wertham (1929) provided the most conclusive data on
episode length with his investigation of 1000 male and
1000 female first admissions for mania. From his published histograms we can today estimate the length of
episodes as lognormally distributed, with a mode between
2– 4 months and a median duration of 4 – 6 months (Q 1 5
2– 4 months, Q 3 5 8 –10 months). There was no
follow-up; most of these manic cases may therefore have
been bipolars.
Rennie (1942) found that there was a lengthening of
repeated manic episodes, with the first episode lasting 3.5
months, the second 5.2 months, the third 5.8 months, and
the fourth and fifth even longer. Rennie ascribed this to a
prolonging effect of aging, since before the age of 45
episode length remained constant.
Unlike Rennie, Kinkelin (1954), in a longitudinal study,
found no systematic change in episode length between the
first and seventh episodes. Among 347 bipolar patients the
mean length of depressive episodes varied impressively
from 3.5 to 8.4 months, and of manic episodes from 4 to
11.6 months. Kinkelin, too, concluded that later episodes
tended to be longer.
In the NIMH Collaborative Study on the Psychobiology
of Depression Clinical Studies (Keller et al 1986a) sur-
vival analysis demonstrated that mixed or cycling episodes
were slower to recover than pure depressed or pure manic
episodes. Perugi et al (1997) also reported that episodes
involving mixed states lasted significantly longer (mean
13.4 months) than manic episodes (8.8 months).
The computation of the length of episodes has to take
into account their lognormal distribution (Angst and Weis
1967) and to control for the individual number of episodes
(Slater 1938). In patients experiencing multiple cyclic
(bipolar) episodes, the episodes tend to be slightly shorter,
whereas in general the latest episode tends to be longer
because of the development of chronicity in some cases, as
found in our multicenter study on the course of mood
disorders (Angst et al 1973).
In the Baltimore Epidemiologic Catchment Area Follow-Up Eaton et al (1997) found a median length of
episodes of approximately 8 to 12 weeks; this is shorter
than that reported in treated populations, but similar to the
median episode length of 8 weeks reported in a community study of adolescents in Oregon (Lewinsohn et al
1994).
The most recent data from the Zurich follow-up study
(Angst and Preisig 1995a) showed a mean episode length
of 4.3 months (s 5 5.44) calculated on the basis of
intraindividual means. The median length of episodes is
not obviously dependent on the total number of episodes—see Table 2, where subgroups with two to 10
episodes are computed separately (following Slater’s
[1938] suggestion). On the whole, the median length of
episodes in bipolar illness (on the basis of individual
medians) was 3 months (Q 1 5 2 months, Q 3 5 5
months). There is a difference in length dependent on
psychopathology. Pure manic and pure depressive episodes lasted 3 months (Q 1 5 2, Q 3 5 5), as did mixed
episodes (M 5 3; Q 1 5 2, Q 3 5 7); in contrast, cyclic
episodes lasted almost 50% longer (M 5 4.19; Q 1 5
2.5, Q 3 5 7.75). These figures are based on intraindividual medians. We found no gender differences in median episode length.
The shorter episode length reported by modern studies
may be a result of their including milder cases or, what is
more probable, a consequence of antidepressant therapy:
medication has usually to be maintained for a further 6
months after recovery to avoid relapse into the still
persistent latent episode, which represents the natural
history.
Recurrence of Bipolar Disorder
The total number of episodes experienced by patients is
amply reported. Findings regarding the occurrence of
single episode cases vary widely, ranging from 0% to 55%
Bipolar Disorder: Natural History
BIOL PSYCHIATRY
2000;48:445– 457
449
Table 2. Total Number of Episodes and Length (Zurich Study)
Episodes
2
3
4
5
6
7
8
9
10
Episodes
Patients
4
8
9
15
14
21
12
13
15
Median length of episodes (months)
1.5
6.0
3.0
5.5
2.5
3.0
2.9
3.5
4.5
5.8
2.8
5.0
4.5
5.5
5.0
3.0
4.5
3.5
9.0
6.0
4.6
5.0
4.0
2.0
4.0
3.0
9.0
4.0
3.8
4.3
3.0
4.0
3.5
5.8
2.8
3.5
2.4
3.0
4.5
2.9
3.0
2.5
2.5
3.0
4.0
4.0
2.5
3.0
4.0
2.0
3.0
1.0
3.0
1
2
3
4
5
6
7
8
9
(Goodwin and Jamison 1990), with five of 11 studies
reporting rates between 0% and 8%. It is clear that any loss
of information will inflate nonrecurrence rates; furthermore, the length of follow-up and development into
chronicity have to be taken into account in assessing
findings.
The earlier literature assumed that periodic mania had a
better prognosis with fewer episodes and better outcome
than periodic melancholia or circular disorder, the latter
having the worst outcome (Rehm 1919, 107); however,
Rehm also found twice as many manic cases (53%) than
circular cases (26%) with short free intervals between
episodes (up to 1 year), reflecting a higher periodicity of
mania (it is unclear whether the length of observation in
these cases was the same).
Certain methodological advances contributed to the
investigation of the question of recurrence. One was the
introduction of a more precise terminology and clear
definitions of remission, recovery, relapse, and recurrence
(Frank et al 1991). Another was the introduction of life
table analyses into psychiatry, a method first used for
length of hospitalization (Kramer 1969) and for follow-up
data after recovery from depression (Klerman et al 1974)
and soon also applied in longitudinal studies of bipolar
disorder (Dunner et al 1976, 1979; Fleiss et al 1976, 1978).
Lavori et al (1984) applied life table methods to reanalyze
40 earlier studies on the relapse/recurrence of affective
disorders; the results varied considerably between the
studies, and the authors formulated the hypothesis of the
heterogeneity of patients’ courses in terms of low or high
hazard with a low or high risk of relapse. More recently,
survival analysis was applied to prospective data on
recurrence after remission (Fleiss et al 1978; Gitlin et al
1995; Keller et al 1993; Lavori et al 1984).
In a 2-year placebo-controlled lithium study Fleiss et al
(1978) found that under placebo 80% of bipolar I patients
experienced recurrences within about 70 weeks. (Admittedly a treatment study of this type dealt with a selected
sample of patients, who, for instance, had had a mean
6.0
10
number of two episodes in the previous 2 years.) The
4-year follow-up of a naturalistic study of mania showed
recurrence in 72% of patients, with a mere 28% remaining
in remission (Tohen et al 1990). Patients who were
symptomatic at the 6-month follow-up had a 45% greater
chance of a recurrence within the next 3.5 years.
Course of Subtypes of Bipolar Disorder
Important findings from a naturalistic study, the National
Institute of Mental Health (NIMH) Collaborative Program
on the Psychobiology of Depression, Clinical Studies
(Keller et al 1993), showed a high rate of recurrence for
pure mania (48% by 1 year and 81% by 5 years) and even
higher rates for the mixed cycling group (57% by 1 year
and 91% by 5 years). Over 7 years the rate of recurrence
was 81%. The length of sustained recovery was associated
with a lower risk for recurrence over the subsequent 4
years, but over a period of 10 years this predictive power
decreased considerably (Coryell et al 1995b); the authors
showed that even under sustained lithium prophylaxis
recurrences were present in more than 70% of cases within
5 years of recovery. This finding is consistent with the
outpatient study of Gitlin et al (1995), in which 73% of 82
bipolar patients had relapses/recurrences over an average
of 4.3 years despite maintenance pharmacotherapy (two
thirds of patients who relapsed experienced multiple relapses). Among the 26 patients who suffered no relapse
46% continued to show significant symptoms of mania or
depression.
The NIMH Collaborative Program on the Psychobiology of Depression, Clinical Studies (Coryell et al 1989)
provided no evidence that the course of bipolar II disorder
differed from that of bipolar I, which confirms our own
findings (Angst 1986).
A recent study comparing patients suffering from bipolar II disorder with major depressive subjects under
fluoxetine showed similar relapse/recurrence rates of 36%
versus 35% after 50 weeks and 44% versus 49% after 62
450
BIOL PSYCHIATRY
2000;48:445– 457
weeks (Amsterdam et al 1998). This study provides no
indication of a deterioration in the course of bipolar II
disorder due to antidepressant medication; on the contrary,
the disorder had a better course than under placebos.
Marneros et al (1991a) reported that bipolar disorder
and schizo-bipolar disorder had very similar course
characteristics.
Is Recurrence of Bipolar Illness
Progressive?
Describing the intervals between episodes, Kraepelin
(1913, 1365) found a progressive shortening of the first
three intervals (first interval of male/female subjects 4.6/
4.3 years, second interval 2.8/2.0 years, third interval
1.2/1.4 years). His data, like that of some other authors, are
not controlled for the total number of episodes: clearly the
intervals between episodes in patients having few episodes
over a lifetime will be longer than in patients who
experience many. We owe this particular methodological
breakthrough to Eliot Slater (1938), who, in his paper on
the periodicity of manic– depressive insanity, investigated
116 patients of the Forschungsanstalt für Psychiatrie in
Munich who had been personally diagnosed by Kraepelin.
Studying the length of intervals between episodes, Slater
made his major methodological contribution by controlling for the total number of episodes and analyzing
separately the subgroups with 1, 2, 3, . . . , n intervals. He
showed that there was indeed a shortening of the intervals,
but only between the first few manifestations of the illness.
Investigating the individual periodicity, he concluded that
every patient had his or her own rhythm.
In 1968 Bratfos and Haug applied Slater’s method to the
follow-up data on 215 cases of manic– depressive disorder
(including depression) in an analysis of the length of the
intervals between episodes. Correcting for the number of
episodes, the authors found that the first interval length
was 2.1 years, the second was 4.8 years, and the third was
2.2 years; thus no clear tendency emerged from these data.
Again following Slater’s method of correcting for the
number of episodes, Angst et al (1973) published the
results of a multinational retrospective and partially prospective hospital record study on the course of 393 bipolar
and 634 unipolar depressive patients. Analyzing cycle
lengths, they indirectly confirmed Slater’s finding that the
intervals shortened as the number of recurrences increased; however, the median cycle lengths gave clear
evidence only for continuous shortening of the first three
cycles; the pattern of later recurrences seemed to be
unpredictable. On the other hand, nonparametric tests
between successive cycle lengths showed a significant
shortening from cycle one to cycle 11 (Angst et al 1973,
499). Marneros et al (1991a) confirmed the shortening of
J. Angst and R. Sellaro
cycles on the basis of the Cologne naturalistic follow-up
study of 30 bipolar and 56 schizo-bipolar patients. Zis et al
(1980) and Zis and Goodwin (1979) arrived at similar
conclusions. This seemed to confirm earlier findings that
had suggested decreasing cycle length (e.g., those of
Kraepelin [1913, 1325] or Kinkelin [1954], although not
controlled for number of cycles—see Figure 6-3 in Goodwin and Jamison [1990]). All these results, together with
the finding that precipitation rates decrease (Angst 1966,
41) with increasing recurrence, led Roy-Byrne et al (1985)
to speak of sensitization and Post et al (Post 1992; Post et
al 1984, 1986) to develop the theory that vulnerability
grows with the number of episodes and the theory of
conditioning, sensitization, and kindling (by analogy with
electrophysiologic kindling).
National data on hospital admissions and readmissions
can also provide a rough estimate of the natural history of
severe cases. Kessing et al (1998a, 1998b) recently described the course of Danish hospital admissions on the
basis of a nationwide register of ICD-8 diagnoses. Rehospitalizations were taken as a measure of recurrence. The
authors found, on the basis of 2903 bipolar cases, a
progressive shortening of the interval between discharge
from hospital and the next rehospitalization and, therefore,
a deteriorating course. Selection bias did not completely
explain the shortening intervals between hospital admissions, and control for gender and age did not alter the
conclusions. So far, then, it would seem to have been
established that the course of bipolar disorder is recurrent
and progressive, but this aspect is still surrounded by
considerable controversy, which may partially be due to
the possibility “that recent studies deal with different,
more broadly diagnosed populations than the seminal,
earlier studies” (Grof et al 1995).
The Lithium Controversy: Does Recurrence
Improve Spontaneously?
One controversy about the natural history of bipolar
disorder dates back to 1968 and the criticism by two
reputed British authors, Blackwell and Shepherd of lithium trials. They assumed not only the inefficacy of lithium
but also that bipolar disorder had a good prognosis,
making long-term prophylaxis unnecessary. This view was
supported theoretically by Lader (1968) and empirically
by Saran (1969), who found no evidence for high recurrence in his follow-up data and who concluded that past
recurrence was not predictive for recurrence in the future.
Saran’s findings were of critical importance because the
early work on lithium had been based statistically on the
assumption that high recurrence in a patient’s previous
history should be expected to repeat itself in the future.
Saran concluded that his findings on the spontaneous
Bipolar Disorder: Natural History
course corresponded to the course as observed under
lithium treatment by Baastrup and Schou (1968). Saran’s
findings have not been confirmed by other studies: a
methodological investigation by Isaaksson et al (1969)
demonstrated the persistence of recurrence for bipolar and
unipolar depression, and the same conclusion was reached
by Laurell and Ottosson (1968).
Stability of Recurrence after Initial
Deterioration
Another debated question is whether the course of bipolar
disorder is really progressive, characterized by unlimited
shortening of cycles throughout (Angst et al 1973). This
model is disconfirmed by a number of new studies. The
recent summary of the literature by Kessing et al (1998b)
shows that Fukuda et al (1983), who investigated not the
early but the later course of the illness, could find no
shortening of cycles (a finding that would still be compatible with the hypothesis that shortening is a feature of the
first few cycles only). This aspect of the systematic
shortening of cycle length was also very seriously questioned in the reviews by Coryell and Winokur (1992) and
Solomon et al (1995), which drew mainly on Winokur and
colleagues’ findings (1994) from the prospective naturalistic NIMH study covering 10 years (Turvey et al 1999).
This study found that the second cycle was clearly shorter
than the first, the third a little longer than the second, but
the fourth and fifth cycles were again shorter. On the other
hand, Winokur et al (1993) stressed that bipolar illness
was highly recurrent, with an “inexorable continuation of
episodes and hospitalisations,” and could find no data
suggesting that the illness burned out at a later stage.
On the basis of prospective data Coryell and Winokur
(1992) found that rapid cycling, which is observed in 20%
to 25% of bipolar patients, is usually a transient manifestation and not therefore a characteristic of the long-term
course.
Meta-Analysis of Two Studies on the Course
Samples and Methodology
Here we reanalyze some data from our two studies, the
mainly retrospective multicenter study published by Angst
et al (1973) and the Zurich follow-up study described by
Angst and Preisig (1995a).
1. The multicenter study (Angst et al 1968a, 1973)
consisted of consecutive hospital admissions of
bipolar and unipolar depressive patients from Basel
(103); Berlin (104); Hamilton, Canada (69); Prague
(132); Zurich (392); Glostrup, Denmark (100); and
Landeck, Germany (140). For the present analysis
BIOL PSYCHIATRY
2000;48:445– 457
451
the Zurich sample was excluded, to avoid any
overlap with the Zurich follow-up study. The remaining sample consisted of 329 bipolar patients.
The data collection was mainly of the retrospective
type based on case histories and verbal information.
The documentation of the course was carried out by
means of a standardized form of protocol in which
all data were entered separately for each episode and
subsequent interval (Angst and Weis 1968). In
retrospect the onset (date) and length (months,
weeks) of previous episodes and aspects of treatment (none, ambulatory, hospital) were assessed.
The degree of remission was coded as full, partial,
or unknown. Psychopathology was coded with a list
of 10 syndromes. The data were reanalyzed for this
study.
2. The Zurich follow-up study consisted of 406 consecutive hospital admissions for severe depression
or mania from 1959 to 1963. Regular follow-up
investigations (by telephone, interviews, and record
collection) were conducted in 1963, 1965, 1979,
1975, 1980, and 1985. In 1991 and 1997 mortality
data were available from the Swiss federal office of
statistics. Seventy-six percent of patients had died
by the end of 1997. The sample and the methodology of assessments were described in detail by
Angst and Preisig (1995a). The principal data collected were comparable to the multicenter study, but
psychopathology and treatment were assessed in
more detail. The 220 bipolar patients were reanalyzed for this article.
Results
Table 3 presents the data from the mainly retrospective
multicenter study and demonstrates a systematic shortening of the first four cycles. The predominantly prospective
data from the Zurich sample (Table 4) shows a significant
shortening between cycles 1 and 2 only, with no systematic change thereafter. In both studies the conclusions were
confirmed by t tests for dependent measures.
In a survival analysis of the Zurich follow-up data
significant differences were found between cycles 1–5 but
were difficult to interpret. The first cycle was longer and
the second cycle shorter than all the others. Otherwise the
survival curves were very similar (Figure 3). The mean
cycle length was 28.7 months (s 5 30.93) but the
median length, which is much more representative, only
18 months (Q 1 5 3, Q 3 5 18); exclusion of the first
longer cycle does not change this median.
In conclusion, we found a shortening of cycle length at
the beginning of the disorder only; later episodes were
persistently recurrent but occurred at irregular intervals
without any systematic deterioration or amelioration, thus
452
J. Angst and R. Sellaro
BIOL PSYCHIATRY
2000;48:445– 457
Table 3. Total Number of Cycles and Length (Multicenter Study)
Cycles
1
2
3
4
5
6
7
8
9
Patients
18
35
38
45
39
31
26
27
17
Cycles
Median length of cycles (months)
23
26
38
21
35
42
28
42
37
15
20
27
18
31
18
17
20
15
13
19
19
16
14
18
14
13
15
12
14
9
11
15
13
9
8
9
15
11
12
12
9
15
10
15
12
1
2
3
4
5
6
7
8
9
This table does not include Zurich data.
confirming Winokur and colleagues’ (1993, 1994)
findings.
Episode Frequency per Year
Periodicity can also be expressed by the episode frequency
per year. In a follow-up of 140 bipolar I patients over 11.4
years Dunner et al (1979) found 0.54 episodes per year; we
found 0.37 (Angst and Preisig 1995a). Marneros’ group
found 0.41 episodes per year (Marneros 1999; Marneros et
al 1991a) for bipolar disorder and about half as many in
the case of recurrent depression. These figures included
the first cycle, which is considerably longer than later
ones. Dunner et al (1979) found no relationship between
episode frequency and age of onset (which was not
consistent with our early results; Angst and Weis 1967).
The question of a gender difference in episode frequency
per year remained open. After the onset of their disorder,
bipolar patients spent on the average about 19% of their
lives in affective episodes over an observation period of 27
years (Angst and Preisig 1995a).
In the present analysis the median cycle length (on the
basis of individual medians) was 18 months (Q 1 5 12
months, Q 3 5 33.5 months), a length that corresponds to
a recurrence rate of 0.66 episodes per year. On the basis of
an episode length of 3 months (median), bipolar patients
spent about 2 months/year in episodes. We could not find
any gender differences (Mann–Whitney U test) in episode
frequency measured by cycle length. Recently Gottschalk
et al (1995) found some evidence for a nonlinear deterministic structure in long-term daily mood records of six
out of seven bipolar patients, something not found in
normal control subjects. Further research on a chaotic
course is certainly desirable.
Correlates of Recurrence
For the analysis of correlates with long-term recurrence
we excluded the first two cycles and entered the median of
all intraindividual medians of cycle length as the dependent variable into a multiple regression analysis. Cycle
length did not correlate with gender, retrospective versus
prospective data collection, or long-term administration of
tricyclic antidepressants. This finding does not confirm the
assumed deterioration of the course of bipolar illness
under tricyclics (Arnold and Kryspin-Exner 1965; Koukopoulos et al 1980; Till and Vuckovik 1970; Tondo et al
1981). Unsurprisingly, long-term lithium and antipsy-
Table 4. Total Number of Cycles and Length (Zurich Study)
Cycles
1
2
3
4
5
6
7
8
91
Cycles
Patients
4
8
9
15
14
21
12
13
124
Median length of cycles (months)
64
105
55
78
30
44
42
34
26
12
21
25
22
35
23
22
19
34
33
37
36
34
23
21
45
19
24
16
30
15
20
24
17
23
15
17
30
12
12
17
17
12
28
12
1
2
3
4
5
6
7
8
12
91
Bipolar Disorder: Natural History
BIOL PSYCHIATRY
2000;48:445– 457
453
Figure 3. Survival analysis of cycle lengths (first to
fifth).
chotic medication correlated with shorter cycles; this
result is explicable by the selection of highly recurrent
cases for prophylaxis. The syndromal characteristics of
episodes had some impact on cycle length: it was significantly shorter in cases with cyclic versus pure (manic or
depressive) episodes and in those with more manic than
depressive episodes.
Poor Outcome of Bipolar Disorder in Early
Follow-Up Studies
In the predrug era most bipolar cases were described as
manifesting residual symptoms after recovery from episodes. Their state was described as “unsteady, moody,
irritable, indolent, egocentric” (Pilcz 1901, 61– 62), a
symptom that often preceded the episodes or had even
been present since childhood. Kraepelin (1913, 1185), too,
although assuming a good prognosis, admitted the existence of mild residual “debility states.” Ziehen (1902)
estimated complete recovery in periodic mania to occur at
most in 20% of cases.
In a follow-up of first admissions from 1920 to 1947
over a mean of 22 years Kinkelin (1954) found chronicity
(including severe residual states) in 14.6% of depressive
subjects (N 5 89) and in 41% of circular cases (N 5
51).
An important follow-up study of 297 bipolar and 945
unipolar patients in the Phipps Clinic (admissions from
1913 to 1940) was published by Stephens and McHugh
(1991). Compared with depressive subjects, bipolar patients had an earlier age of onset on admission and were
more likely to be psychotic; they also had more serious
premorbid characteristics, more sudden onsets (44%),
more previous admissions (62% for depression and 67%
for mania), and more problems with alcohol, but more
bipolar than depressed patients were discharged in a
recovered state. In bipolar disorders they found a lifetime
average of 3.6 hospitalizations, and in their follow-up they
observed 0.19 episodes of depression and 0.29 episodes of
mania per year followed up. Only 2% of all patients had
manic episodes with no lifetime depressions; this small
group had the best outcome. Bipolar patients (N 5 301)
had the poorest outcome: 7% were recovered, 50% improved, and 43% unimproved at the follow-up rating. The
rates of recovered, improved, and unimproved patients
were 19%, 56%, and 25% for pure mania (N 5 16) and
25%, 44%, and 32% for pure depression (N 5 700).
In a follow-up of 86 manic patients over 30 to 40 years,
outcome (measured by psychiatric symptoms) was good in
43% of cases, fair in 18%, and poor in 25% (Tsuang et al
1979).
Poor Outcome in Recent Follow-Up Studies
Modern studies (e.g., a 4.5-year follow-up study by
Goldberg et al [1995]) showed that only 41% of bipolar
patients had a good overall outcome, the remainder being
moderately impaired (37%) or showing poor functioning
(22%). Gitlin et al (1995) found a poor outcome even
under prophylactic medication (mainly lithium); poor
outcome was more closely associated with the number of
depressive episodes than the number of manic episodes. In
their above-mentioned NIMH Collaborative Program on
the Psychobiology of Depression with a 15-year intensive
follow-up, Coryell et al (1998) found that 56.6% of 113
bipolar I patients had had no symptoms over the past 12
months, 22.1% had had symptoms for fewer than 52
weeks, and 20.4% had had symptoms in all 52 weeks,
representing a poor outcome (defined as having had
symptoms of major depressive disorder, mania, or schizoaffective disorder in all 52 weeks of year 15).
The short-term outcome, 12 months after discharge
from hospital, was similar in patients with manic episodes
or mixed episodes (Keck et al 1998), whereas in Keller
and colleagues’ (1986b) study after 18 months of follow-up mixed episodes also developed more frequently
454
J. Angst and R. Sellaro
BIOL PSYCHIATRY
2000;48:445– 457
Table 5. Outcome of 219 Patients in the Zurich Follow-Up
Study (Median Age at Follow-Up or Death, 68 Years)
Recovered (GAS score . 60, no episodes over the past
5 years)
Remitted (GAS score . 60) but still recurrent (,5
years since last episode)
Incomplete remission (GAS score 1– 60) over more
than 5 years
Incomplete remission, course still recurrent
Chronic (last episode without remission, minimum
length 2 years)
Suicide
16.0%
25.5%
2.
7.8%
27.0%
15.9%
3.
7.8%
GAS, Global Assessment Scale.
into chronicity (32%) than pure manic episodes (7%). In
bipolar I patients poor outcome did not correlate clearly
with early onset of the disorder or cycling per se (Coryell
et al 1998), but persistence of depressive symptoms in
years 1 and 2 was correlated with impairment after 15
years (household duties, recreational activities, overall
satisfaction, and global social adjustment). Such early
persistence of depressive symptoms predicted a poor
prognosis of a bipolar subtype, whereas this was not the
case for early persistence of manic symptoms.
The lifetime outcome of bipolar disorder in our Zurich
follow-up study is given in Table 5 and demonstrates a
poor prognosis despite modern treatments. Up to a median
age of 68 years only 16% of patients had recovered; 52%
still suffered from recurrent episodes and the remaining
patients had become chronically ill or had committed
suicide. These data underline the poor outcome into old
age and the need for intensive treatment. This table shows
the outcome of bipolar disorder before the occurrence of
an organic brain syndrome in the elderly, which was found
in 14.5% of cases. The previous number of affective
episodes was not correlated with the development of an
organic brain syndrome (Angst and Preisig 1995b).
Future studies on outcome should clearly distinguish
between different outcome measures, since Tohen et al
(2000) have shown that functional recovery can be much
worse than syndromal recovery.
Conclusions
Several important conclusions regarding episodes, recurrence, and outcome have emerged from this review and
data analysis of the natural history of bipolar disorder:
1. Before the introduction of modern drugs, spontaneous mild depression following a manic episode and
spontaneous hypomania following a melancholic
episode were very common; they were interpreted as
“reactive” and had no effect on the principal diagnosis of pure mania or pure melancholia. This
4.
5.
6.
7.
historical fact has to be taken into account in today’s
hypothesis of drug-induced hypomania or druginduced depression; these have to be proven statistically by placebo-controlled trials.
Our decades-long prospective study shows that over
lifetime the proportions of mania and depression in
bipolar disorder remain stable into old age. Bipolar
female subjects manifest more depression than bipolar male subjects.
Mixed states have been described since the early
19th century. Modern studies have demonstrated
that they have a poorer prognosis than other bipolar
conditions, with slower remissions and higher risk
for chronicity.
The natural length of affective episodes has probably not changed over the past 120 years. Patients
responding to antidepressants still require a maintenance treatment throughout the underlying episode.
In clinical studies the median length of episodes is 3
to 6 months; in epidemiological studies it is 2 to 3
months.
The recurrence of bipolar disorder was always the
rule; it now seems to be established that there is
some initial shortening of intervals/cycles, followed
by an irregular persistent recurrence, with a median
cycling of 18 months. In contrast to earlier reports,
the new studies show that there is no unlimited
shortening of cycle length—not therefore supporting
the kindling model. In several studies rapid cycling
has been found to be relatively frequent, but usually
transient.
Lifelong outcome has rarely been studied, and
precise data on the natural outcome are scarce. Some
new prospective studies demonstrate that most patients continue to suffer from residual depressive or
hypomanic symptoms between episodes, and many
are functionally impaired.
Overall research into the natural history of bipolar
illness shows that it has a poor prognosis, as
reflected by high recurrence, chronicity of episodes
or residual symptoms, and premature death by
suicide and somatic disorders; however, unlike
schizophrenia, which is characterized by much
higher chronicity and the predominance of negative
and psychotic symptoms, chronicity in bipolar disorders is rarer (10 –20%) and the more frequent
residual states are limited to characteristic depressive and hypomanic symptoms.
The authors thank Professor Andreas Marneros (University of Halle) and
Dr. Mauricio Tohen (Lilly Company) for their contribution to this article
in the form of comments and suggestions.
Bipolar Disorder: Natural History
Aspects of this work were presented at the conference “Bipolar
Disorder: From Pre-Clinical to Clinical, Facing the New Millennium,”
January 19 –21, 2000, Scottsdale, Arizona. The conference was sponsored by the Society of Biological Psychiatry through an unrestricted
educational grant provided by Eli Lilly and Company.
References
Amsterdam JD, Garcia-Espana F, Fawcett J, Quitkin FM, Reimherr FW, Rosenbaum JF, et al (1998): Efficacy and safety of
fluoxetine in treating bipolar II major depressive episode.
J Clin Psychopharmacol 18:435– 440.
Angst J (1966): Zur Aetiologie und Nosologie Endogener Depressiver Psychosen. Eine Genetische, Soziologische und
Klinische Studie. Berlin: Springer.
Angst J (1978): The course of affective disorders. II. Typology of
bipolar manic-depressive illness. Arch Psychiatr Nervenkr
226:65–73.
Angst J (1986): The course of major depression, atypical bipolar
disorder, and bipolar disorder. In: Hippius H, Klerman GL,
Matussek N, editors. New Results in Depression Research.
Berlin: Springer, 26 –35.
Angst J (1987): Switch from depression to mania, or from mania
to depression. J Psychopharmacol 1:13–19.
Angst J (1988): Clinical course of affective disorders. In:
Helgason T, Daly RJ, editors. Depressive Illness: Prediction
of Course and Outcome. Berlin: Springer, 1– 48.
Angst J, Baastrup PC, Grof P, Hippius H, Poeldinger W, Weis P
(1973): The course of monopolar depression and bipolar
psychoses. Psychiatr Neurol Neurochir 76:489 –500.
Angst J, Battegay R, Bente D, Berner P, Broeren W, Cornu F, et
al (1968a): Das Dokumentationssystem der Arbeitsgemeinschaft für Methodik und Dokumentation in der Psychiatrie
(AMP). Arzneimittelforschung 18:3– 8.
Angst J, Grof P, Hippius H, Poeldinger W, Weis P (1968b): La
psychose maniaco-depressive est-elle periodique ou intermittente? Dans: Cycles biologiques et psychiatrie. In: de Ajuriaguerra J, editor. Symposium Bel-Air III, Geneve 1967.
Geneva: Georg et Cie, 339 –351.
Angst J, Preisig M (1995a): Course of a clinical cohort of
unipolar, bipolar and schizoaffective patients. Results of a
prospective study from 1959 to 1985. Schweiz Arch Neurol
Psychiatr 146:5–16.
Angst J, Preisig M (1995b): Outcome of a clinical cohort of
unipolar, bipolar and schizoaffective patients. Results of a
prospective study from 1959 to 1985. Schweiz Arch Neurol
Psychiatr 146:17–23.
Angst J, Weis P (1967): Periodicity of depressive psychoses. In:
Brill H, Cole JO, Deniker P, Hippius H, Bradley PB, editors.
Neuro-Psychopharmacology. Proceedings of the Fifth International Congress of the Collegium Internationale NeuroPsychopharmacologicum, Washington, D.C. 1966, International Congress Series No 129. Amsterdam: Excerpta
Medica, 703–710.
Angst J, Weis P (1968): Aetiologie und Verlauf endogener
Depressionen. In: Rémy M, editor. Depressionen und ihre
Behandlung. Fortbildungskurse der SGP, Vol 1. Basel, Switzerland: Karger, 8 –16.
Arnold OH, Kryspin-Exner K (1965): Zur Frage der Beeinflus-
BIOL PSYCHIATRY
2000;48:445– 457
455
sung des Verlaufes des manisch-depressiven Krankheitsgeschehens durch Antidepressiva. Wien Med Wochenschr 115:
929 –934.
Baastrup PC, Schou M (1968): Prophylactic lithium. Lancet
I:1419 –1422.
Baillarger J (1854): De la folie à double forme. Ann Med Psychol
6:369 –384.
Ballet G (1903): Traité de Pathologie Mentale. Paris: Doin.
Blackwell B, Shepherd M (1968): Prophylactic lithium: Another
therapeutic myth? An examination of the evidence to date.
Lancet 1(7549):968 –971.
Bourgeois ML, Marneros A (in press): Course and prognosis of
bipolar disorders. In: Marneros A, Angst J, editors. Bipolar
Affective Disorders. Dordrecht, The Netherlands: Kluwer.
Bratfos O, Haug JO (1968): Course of manic-depressive psychosis. A follow-up investigation of 215 patients. Acta Psychiatr
Scand 44:89 –112.
Coryell W, Endicott J, Maser JD, Keller MB, Leon AC, Akiskal
HS (1995a): Long-term stability of polarity disinctions in the
affective disorders. Am J Psychiatry 152:385–390.
Coryell W, Endicott J, Maser JD, Mueller T, Lavori P, Keller M
(1995b): The likelihood of recurrence in bipolar affective
disorder: The importance of episode recency. J Affect Disord
33:201–206.
Coryell W, Keller M, Endicott J, Andreasen N, Clayton P,
Hirschfeld R (1989): Bipolar II illness: Course and outcome
over a five-year period. Psychol Med 19:129 –141.
Coryell W, Turvey C, Endicott J, Leon AC, Mueller T, Solomon
D, Keller M (1998): Bipolar I affective disorder: Predictors of
outcome after 15 years. J Affect Disord 50:109 –116.
Coryell W, Winokur G (1992): Course and outcome. In: Paykel
ES, editor. Handbook of Affective Disorders, 2nd ed. London:
Churchill Livingstone, 89 –108.
Dunner DL, Fleiss JL, Fieve RR (1976): The course of development of mania in patients with recurrent depression. Am J
Psychiatry 133:905–908.
Dunner DL, Murphy D, Stallone F, Fieve RR (1979): Episode
frequency prior to lithium treatment in bipolar manic-depressive patients. Compr Psychiatry 20:511–515.
Eaton WW, Anthony JC, Gallo J, Cai G, Tien A, Romanoski A,
et al (1997): Natural history of Diagnostic Interview Schedule
DSM-IV major depression. The Baltimore Epidemiologic
Catchment area follow-up. Arch Gen Psychiatry 54:993–999.
Falret J (1861): Principes à suivre dans la classification des
maladies mentales. Ann Med Psychol 19:145.
Falret JP (1851): Marche de la folie. Gazette Hopitaux 24:18 –19.
Falret JP (1854): De la folie circulaire ou forme de maladie
mentale caractérisée par l’alternative régulière de la manie et
de la mélancholie. Bull Acad Med (Paris) 19:382.
Fleiss JL, Dunner DL, Stallone F, Fieve RR (1976): The life
table: A method for analyzing longitudinal studies. Arch Gen
Psychiatry 33:107–112.
Fle